Re: Boost.Threads - Open Questions from the Review

From: cop_at_[hidden]
Date: 2001-09-15 22:56:57

--- In boost_at_y..., dmoore_at_a... wrote:
> > --- In boost_at_y..., John Maddock <John_Maddock_at_c...> wrote:
> > >In Semaphore::up the assert:
> > >
> > >     assert(ret || GetLastError() == ERROR_TOO_MANY_POSTS);
> > >
> > >is failing for the semaphore test in thread_test.cpp, it seems as
> > though
> > >GetLastError is returning ERROR_INVALID_PARAMETER rather than
> > >ERROR_TOO_MANY_POSTS on my Win98 box.
> > 
> > I only have access to Win2K machines at this point, and I can not 
> reproduce 
> > this problem nor can I see a reason for it.  I'd really 
appreciate 
> it if 
> > anyone can explain why this is occuring.
> 
> This is unfortunately one of the many quirks of using the "same" 
> Win32 API on two fundamentally different OS kernels (98 vs 2k).  
Wait 
> until you see the differences in the virtual memory systems if you 
> ever tackle a Boost Memory Map class!
> 
> The MSDN docs are silent on the possible error codes returned by 
> ReleaseSemaphore, actually, but differences in return codes do 
exist 
> throughout the synchronization and VM apis.
> 
> Dave

Since GetLastError() might report different errors on different Win32 
platforms, the only safe assumption is probably assert(GetLastError
() != ERROR_SUCCESS).

Also note that calling GetLastError() in DEBUG only code is dangerous 
because GetLastError() is NOT side-effect free! I tracked down a bug 
at work where the following code was mysteriously failing in debug 
builds:

    assert(GetLastError() == ERROR_SUCCESS); // no errors
    if (GetLastError() != ERROR_SUCCESS) // mysterious error appears?!

The first call to GetLastError() in the assert() apparently cleared 
the thread's error code, so the second call to GetLastError() in the 
non-debug code returned a garbage value! I was very surprised. I 
fixed the problem like this:

    DWORD error = GetLastError();
    assert(ERROR_SUCCESS == error);
    if (ERROR_SUCCESS != error) ... 

Chris